You are on page 1of 11

Acta Tropica 151 (2015) 2131

Contents lists available at ScienceDirect

Acta Tropica
journal homepage: www.elsevier.com/locate/actatropica

Agent-based mathematical modeling as a tool for estimating


Trypanosoma cruzi vectorhost contact rates
Kamuela E. Yong a,b,e , Anuj Mubayi a,c , Christopher M. Kribs a,d,
a

Simon A. Levin Mathematical, Computational & Modeling Sciences Center, Arizona State University, Tempe, AZ 85287, USA
School of Mathematical & Statistical Sciences, Arizona State University, Tempe, AZ 85287, USA
School of Mathematical and Natural Sciences, Arizona State University, Phoenix, AZ 85069, USA
d
Mathematics Department, University of Texas at Arlington, Arlington, TX 76019, USA
e
Mathematics/Science Subdivision, University of Hawai'i West O'ahu, Kapolei, HI 96707, USA
b
c

a r t i c l e

i n f o

Article history:
Received 15 May 2015
Received in revised form 25 June 2015
Accepted 29 June 2015
Available online 26 July 2015
Keywords:
Trypanosoma cruzi
Agent-based model
Contact rates
Host irritability

a b s t r a c t
The parasite Trypanosoma cruzi, spread by triatomine vectors, affects over 100 mammalian species
throughout the Americas, including humans, in whom it causes Chagas disease. In the U.S., only a few
autochthonous cases have been documented in humans, but prevalence is high in sylvatic hosts (primarily raccoons in the southeast and woodrats in Texas). The sylvatic transmission of T. cruzi is spread by the
vector species Triatoma sanguisuga and Triatoma gerstaeckeri biting their preferred hosts and thus creating
multiple interacting vectorhost cycles. The goal of this study is to quantify the rate of contacts between
different host and vector species native to Texas using an agent-based model framework. The contact
rates, which represent bites, are required to estimate transmission coefcients, which can be applied to
models of infection dynamics. In addition to quantitative estimates, results conrm host irritability (in
conjunction with host density) and vector starvation thresholds and dispersal as determining factors for
vector density as well as hostvector contact rates.
2015 Elsevier B.V. All rights reserved.

1. Introduction
At the core of infectious disease transmission are contact processes, through which a pathogen is passed from infective to
susceptible individuals, whether members of the same population (directly transmitted infections) or host and vector. Implicit in
such processes are the movements of individuals, which determine
which others are in close enough proximity for the pathogen to be
spread. For vector-borne infections in particular, vector dispersal in
search of hosts on which to feed is the engine that drives the spread
of the disease. For many disease vectors, such as mosquitoes, feeding contacts and dispersal are completely aligned: each bloodmeal
typically involves a new host, since hosts are plentiful. Other vectors, however, nest with individual hosts, only leaving in search of
a new one when the old host dies or fails to return to the nest.
For these vectors, the acts of feeding and dispersal are distinct,
and occur with markedly different frequencies, raising the question of which gives the better measure of disease transmission

Corresponding author at: Department of Mathematics, University of Texas at


Arlington, Box 19408, Arlington, TX 76019-0408, USA.
E-mail address: kribs@uta.edu (C.M. Kribs).
http://dx.doi.org/10.1016/j.actatropica.2015.06.025
0001-706X/ 2015 Elsevier B.V. All rights reserved.

and spread (since repeated contacts between the same host and
vector have limited transmission potential). This is the case with
the triatomine vectors responsible for the spread of the protozoan
parasite Trypanosoma cruzi, the etiological agent of Chagas disease.
T. cruzi, which is native to the Americas from the southern U.S.
down to the southern cone, is carried by numerous species of triatomine vector, most of the genus Triatoma (and each limited to a
particular habitat), and transmitted to over 100 mammalian host
species. Triatomine bugs pass through ve distinct instar stages
before reaching adulthood, requiring at least one bloodmeal at
each stage. Unlike mosquitoes and many other disease vectors, triatomines opportunity to feed is limited by host availability and
irritability (Schoeld, 1982; Schoeld et al., 1986): nymphs typically remain in the nest or den where they hatch, while adults often
disperse in search of new hosts (Guerenstein and Lazzari, 2009;
Noireau and Dujardin, 2010). Primary hosts such as raccoons (Procyon lotor) and opossums (Didelphis spp.) typically have multiple
sleeping places throughout their range and move within that territory to forage, returning to a given den only after several days
(Rabinowitz and Pelton, 1986; Shirer and Fitch, 1970). (Woodrats
move less but have been known to use as many as 5 different nests
separated by hundreds of meters (Merkelz and Kerr, 2002).) In
addition, when too many vectors attempt to feed on a given host,

22

K.E. Yong et al. / Acta Tropica 151 (2015) 2131

that host takes measures to defend itself, including leaving the den
altogether. Since only adult triatomines have wings, and those are
more efcient for gliding than ying, it is easier for a host to defend
itself against them than against insects such as mosquitoes. Indeed,

one modeling study (Castanera


et al., 2003) indicated that host
irritability plays an important role in vector demographics. When
hosts return to a given den or nest often enough, if there are few
enough vectors there, they may all be able to feed often enough to
survive, mature and reproduce, but when a nest overpopulates, or
a host is gone too long, adult triatomines will disperse in search of
another host, making use of chemical, thermal, and even infrared
cues to detect hosts and infested nests (Lazzari et al., 2013). It is the
successful move to a new host, called host switching, that allows T.
cruzi to spread, within and even between host populations.
Several modeling studies have already examined the structure underlying hostvector contacts for T. cruzi, in particular
how the contact rates saturate in host or vector populations, as
a function of the average vector-to-host ratio (Kribs-Zaleta, 2004,
2006, 2009, 2010a,b; Pelosse and Kribs-Zaleta, 2012). (Transmission dynamics are even further complicated by the fact that some
hosts, as opportunistic feeders, sometimes prey on the vectors they
nd nearby, but this study focuses on classical vector-initiated
hostvector contacts.) This saturation can affect vector population
dynamics, T. cruzi transmission dynamics, and even which strain
of T. cruzi is advantaged in a cycle where multiple cocirculating
strains compete for access to hosts. However, answers to all of
these questions require estimates of the basic hostvector contact rates and related quantities, which remain extremely difcult
to measure directly in situ. Laboratory studies have investigated
individual vector species preferred feeding frequencies (and average time to defecation, which is key to gauging the efciency of
classical stercorarian transmission of T. cruzi) but cannot take into
account the real availability of hosts. One notable study included
both lab and eld components, measuring the mating duration
and frequency, egg size, number and weight, laying and hatching
times, hatch proportion and number of ovipositions, developmental periods and longevity of each stage including adults, feeding
preferences on eleven host species including nonmammalian vertebrates, feeding and defecation times, volume and weight of
bloodmeals, time to starvation, intraspecic parasitism, T. cruzi
infection prevalence, nocturnal ight activity levels, and host association of three triatomine species (Triatoma sanguisuga, Triatoma
gerstaeckeri, Rhodnius prolixus), in some cases as a function of temperature (Pippin, 1970). However, it did not capture hostvector
contact frequencies in the eld. (It also revealed some discrepancies
between laboratory and eld-based measures of the same quantities.) A recent study provided a comprehensive review of host and
vector demographics and T. cruzi prevalence in species present in
the U.S. (Kribs-Zaleta, 2010a), but the average hostvector contact
rates could only be back-estimate assuming observed prevalences
represented endemic steady states, and even then the rates incorporated probability of infection given a contact. In cycles where
multiple contact-related transmission avenues exist (e.g., stercorarian and oral), direct estimates of these rates remain key to
understanding T. cruzi transmission.
Given the difculties inherent in direct measurement, mathematical models offer one potential way to estimate contact rates
using the basic behavioral tendencies measured in laboratory studies. Compartmental models have long been used to describe the
collective behaviors of populations using relatively little biological
information; however, in recent years agent-based modeling has
developed as a means of using more detailed information at the
individual level to build up emergent properties of systems comprising many individuals (DeAngelis and Gross, 1992; Grimm and
Railsbeck, 2005). As Gaff wrote (Gaff, 2011), Agent-based models,
also called individual-based models, are computer-based models

that simulate the actions and interactions of autonomous agents


that represent the individuals of the population. The activity of
these agents can be compiled to obtain population-level measures
of a biological system at a given scale. Gaff, for instance, developed an agent-based model (ABM) of deertick interaction (Gaff,
2011) in which tick vectors (Ixodes spp.) passed through four life
stages (egg, larva, nymph, adult), with each transition requiring a
single bloodmeal, and found that the ABM predicted signicantly
lower infection rates than a classical ordinary differential equations
(ODE) model with analogous parameter values, due to the ABMs
discrete, limited host interactions for each vector (contrasted with
the continuous hostvector interactions implied by the ODE system). ABMs have also been used to study T. cruzi transmission to
humans (Chagas disease). A group of French researchers (Barbu
et al., 2010; Slimi et al., 2009) used a spatially explicit cellular
automaton model with discrete individuals to study spatial infestation patterns of a Mexican rural village by the T. cruzi vector
Triatoma dimidiata, nding that the constant inux from the forest surrounding the village made the villages periphery the most
important place to apply vector control (insecticide). Galvo et al.
(2008) used a within-host ABM to study the impact of T. cruzi at
the cellular level on cardiac tissue regeneration following a bone
marrow stem cell transplant. Devillers et al. (2008) used an ABM
to study the competition for hosts between two T. cruzi types in
South America, nding that sylvatic reservoir hosts were necessary in order to explain the observed prevalences in humans. The
present study thus proposes to use known information about the
behaviors of individual sylvatic hosts and vectors to study the basic
contact mechanisms (vector feeding and host switching) responsible for the spread of T. cruzi in sylvatic reservoir cycles, through the
framework of an ABM. Model simulations will use the contexts of
raccoons and (separately) woodrats (Neotoma micropus), common
hosts in the southern U.S., and their associated vectors T. sanguisuga
and T. gerstaeckeri (Kjos et al., 2009), although the model structure
can be applied to other hosts and vector species as well.
The primary research questions are thus: How often do
vectorhost contacts really occur? and what impact does host
irritability have on that rate? (A related question is, how often
does host switching occur within a given cycle?) Underlying these
questions is the more complex issue of how best to use these baseline rates to describe the rate at which T. cruzi actually spreads.
The following section details the model structure, built around the
key elements of host and adult vector movement and demographics, and vector feeding and maturation. Later sections present and
interpret results obtained from simulations; this study will focus
on the relationship among host irritability and vector dispersal and
starvation, and how that relationship drives contact rates.
2. Model
The ABM developed for this study was coded in NetLogo 5.0.3
(Wilensky, 1999). The model description in this section follows
the ODD (Overview, Design concepts, Details) protocol for describing ABMs (Grimm et al., 2006, 2010) using the complete required
format.
2.1. Purpose
The ABM developed for this study uses information from the
scientic literature about the demographics and movement of individual hosts (raccoons or woodrats) and vectors (T. sanguisuga
and/or T. gerstaeckeri), and about the feeding preferences of vectors, to estimate the rates at which vectors feed on hosts and switch
hosts in the eld, as a function of host irritability level (maximum
number of bites tolerated by a host in one night) and transmission

K.E. Yong et al. / Acta Tropica 151 (2015) 2131

cycle (host species). The resulting estimates will inform the current understanding of how the parasite Trypanosoma cruzi spreads
at the population level in sylvatic settings.
2.2. Entities, state variables, and scales
The entities (agents) in the model include hosts (raccoons and
woodrats), vectors (T. sanguisuga and T. gerstaeckeri), and homes
(raccoon dens and woodrat nests). Hosts have counters for the
number of vectors of each species trying to bite them on that time
step (for purposes of determining host irritability the counters are
reset at each time step). Vectors likewise have counters marking
whether or not they try to bite a given host type during the present
time step; hunger counters, marking the number of turns (days)
since last feeding; variables recording the types of patch in which
they currently and most recently resided, for purposes of marking
host type switching across cycles (not implemented in this study);
and two lists describing their personal feeding histories, marking
the time-index and host-ID at each feeding (stored to a global list
at death). Homes have counts of the number of juvenile vectors
of each species living there, as well as a host absence counter,
which also serves as a hunger counter for the juveniles. For simplicity, homes are assumed permanent, created at setup and never
destroyed. All agents also have (x, y) coordinates (for homes these
are xed).
The world consists of a 75 75 grid, with opposite boundaries
identied for continuity (left to right and top to bottom, as is commonly done, but note this creates a torus geometry). Cell diameters
were set at 10.4 m, since this reects the approximate average minimum distance between woodrat nests as well as mean distance
traveled daily by a woodrat (Rogers et al., 2010; Raun, 1966; Thies
et al., 1996); thus the model simulates an area of about 60 ha or
150 acres. Each patch (grid cell) has a landscape property taking
on one of the three possible values (permanent over the course
of a simulation): wooded (raccoon habitat), prickly pear cactus
(woodrat habitat), or neither. For the purpose of this study, simulations were run with all cells of the same landscape type (wooded
for raccoons, or prickly pear for woodrats).
The time step used is one day, since hostvector contacts only
occur when hosts return to their sleeping places to rest after foraging. Each simulation was run for 100 days.
2.3. Process overview and scheduling
Following setup (q.v.), each day the following ve routines occur
in the order given below, until either the specied number of days
passes or there are no more vectors, whichever occurs rst:

hosts (woodrats and raccoons) move;


vectors of each type move (disperse if hungry enough);
host births and deaths;
vector natural death, feeding, and starvation;
homes (nests and dens) are updated: primarily juvenile vector
birth, death and maturation.

2.4. Design concepts


Basic principles: In accordance with research literature, raccoons
are assumed to be solitary and have several sleeping places across
their territory, while woodrats have relatively few and may share
a nest (generally of the other gender, as for breeding, but in this
model gender is not made explicit, and nests may simply hold up
to two woodrats at a time). Juvenile vectors of either species are
assumed incapable of dispersal (since they have no wings and are
commonly found in nests or dens), while adult vectors are assumed
to disperse only to avoid starvation. Hosts are assumed to be able

23

to defend themselves and/or leave a home in order to receive no


more than a given number of bites per night (the host irritability
threshold). Bites by juvenile vectors are assumed to provoke less
irritation than bites by adult vectors, by a given factor.
Emergence: The models primary outputs are measures of vector feeding and host-switching rates, at the population level but
compiled from individuals.
Adaptation: In foraging from day to day, hosts are assumed to
seek (and nd) the least-recently visited site within a days range;
hosts are also assumed able to defend themselves against excessive
bites per night. Adult vectors are assumed to get hungry a certain
number of days after feeding (threshold 1), disperse in search of
a (new) host after another period (threshold 2), and nally starve
if they have not fed after another period (threshold 3); when dispersing, adult vectors are also assumed able to nd occupied homes
within a days travel radius.
Sensing: Hosts are assumed to be able to detect which part(s) of
their home ranges (represented by dens or nests) within a single
days travel have had longest to replenish food sources. Vectors
are assumed to be able to detect their own hunger thresholds, and
dispersing vectors are assumed to be able to detect hosts located
within a single days travel of their present locations.
Interaction: The primary interactions of interest involve vectors
feeding on hosts, within the connes of a den or nest, subject to the
host irritability threshold (thereby making competition for access
to hosts an indirect interaction between vectors). Births of new
hosts or juvenile vectors require the presence of an adult of the
given species, an indirect interaction.
Stochasticity: In setup (q.v.), the spatial conguration of landscape types (but not the numbers of cells of each landscape type)
is random, as are which specic nests and/or dens are infested
with vectors (but not the numbers of each that are infested),
which nests/dens serve as initial locations for the hosts, and which
infested nests/dens serve as initial locations for the adult vectors.
The initial value for each homes host absence counter is taken
from a uniform distribution bounded above by the vector dispersal
threshold. Initial numbers of juvenile vectors of each species in
infested homes are drawn from exponential distributions with the
corresponding expected means.
During runtime, juvenile vector births, survival, and maturation,
adult vector survival, and host birth and survival are all taken from
binomial distributions; once the numbers are determined, which
host or adult vector dies is random. Newborn hosts are placed
randomly in a home of the appropriate type with room for them.
Host movement (from one home to another), when not triggered
by host irritability (excessive bites), is determined stochastically
with a given probability, and when either (a) there are multiple
homes within a days travel for that host with maximal host absence
counter, or (b) there are no unoccupied homes within a days travel
but there are homes with room for the given host, the home to
which the host moves is chosen randomly. When an adult vector
disperses without reaching an occupied home immediately, its distance traveled in a day is drawn from an exponential distribution,
and its direction is random. When too many vectors attempt to
bite a given host, which adult vectors are denied feeding is chosen
randomly.
Observation: At the end of each simulation, the collected individual feeding histories (timestamp and host ID for each successful
feeding, as well as timestamps of birth and death, from which one
can determine feeding and host-switching rates) of all adult vectors
are stored for analysis, as well as the collected dispersal results. In
addition, NetLogos Behavior Space allows one to tabulate the number(s) of agents in various states at each time step. In this study
the numbers of vector bites, hosts, and vectors were recorded at
each time step in order to calculate total instantaneous per-host or
per-vector contact rates.

24

K.E. Yong et al. / Acta Tropica 151 (2015) 2131

2.5. Initialization
Setup begins by assigning each grid cell a landscape type
(wooded raccoon habitat, prickly pear cactus woodrat habitat, or
neither): each of the two habitat types has a preset percentage of
cells, and the corresponding number of cells is then chosen randomly from the grid and assigned. Next appropriate numbers of
woodrat nests and raccoon dens are created, based in each case on
the number of homes per host and number of hosts, of the given
type. Each home is randomly placed on a grid cell of the appropriate habitat type not already occupied by a home. Since homes will
later be seeded with vectors, each homes host absence counter is
assigned by a draw from a uniform distribution on [0, th ], where th is
the vector dispersal hunger threshold (if the host absence counter
exceeded th , the adult vectors would all have left).
Next hosts are created (raccoons and woodrats), each according
to the preset densities (and habitat size), and placed randomly in
unoccupied homes (woodrats are allowed to be placed up to 2 in
a nest, as with breeding pairs although no genders are assigned
explicitly). Their bite counters are initialized to 0.
Homes are then infested with juvenile vectors according to preset infestation percentages for each home type. The number of
juvenile T. sanguisuga in an infested nest is drawn from an exponential distribution whose mean is calculated as follows: The given
per-area vector density is multiplied by the area of the total grid
(in m2 ), scaled down by the [preset, initial] proportion of vectors
which are juveniles, by the proportion of the grid which consists of
woodrat habitat, and by the [preset, initial] proportion of vectors in
a nest which are T. sanguisuga. The numbers of juvenile T. gerstaeckeri in an infested nest, and of juvenile T. sanguisuga in an infested
den, are calculated similarly. Analogous (but simpler) calculations
yield the [deterministic] numbers of adult vectors of each species
in the entire grid; the corresponding adult vectors are then created
and randomly placed in homes already infested with juveniles of
the given species, once their counters and lists are initialized (to 0).
Their patch types are also then set to mark their initial host species
afliation.
Finally, the global result lists that will store dispersal outcomes
and vector feeding histories are initialized (as empty).
2.6. Input data
The model does not use input data to represent time-varying
processes. However, in order to explore the effect of the host
irritability threshold on vector feeding rates, host irritability was
varied across experiments (each experiment consisting of 100 simulations using identical parameter values), set at either 2 or 10 bites
per night. In addition, one set of experiments was run with raccoons
as the only hosts and T. sanguisuga as the only vectors, while a different set of experiments involved woodrat hosts only, and both
species of vector (since both are found in woodrat nests).
Model parameters, available as user inputs in interactive mode,
are summarized in Table 1 and addressed in the following subsection.
2.7. Submodels
1. Move hosts
(a) Move woodrats
Each woodrat is polled in turn to determine whether the
total number of attempted vector bites (calculated as the
sum of three quantities: the number of adult T. sanguisuga
in the present nest which wish to feed as evidenced by their
woodrat bite counters [being positive], the number of adult T.
gerstaeckeri in the nest which wish to feed, and the total number of juvenile vectors in the nest, rescaled both for relative

irritation and desired feeding frequency) meets or exceeds


the host irritability threshold. If not, then a random draw
(from a uniform distribution on the unit interval) is compared with the set probability of switching nests (if less, host
dispersal should occur regardless of irritability). When either
of the two trigger conditions is met, if any unoccupied nests
are within a days travel, the woodrat moves to a randomly
chosen one from among those longest-unoccupied (i.e., with
greatest host absence counter). Otherwise, when dispersing,
the woodrat moves to any randomly chosen nest with no
more than 1 woodrat presently occupying it.
(b) Move raccoons
Raccoons decide whether, and where, to move according to
the same algorithm as for woodrats, except that dens have no
T. gerstaeckeri, and raccoons have different travel distances
and probabilities than woodrats.
2. Move vectors
(a) Move T. gerstaeckeri
Each adult vector, in turn, disperses if its hunger counter
meets or exceeds the dispersal hunger threshold. If any occupied nests are within a days travel (distance drawn from
an exponential distribution with the given mean) the vector enters the nearest one. In this case, the vectors previous
and current patch types, and the global list of dispersal outcomes, are updated. If no occupied nests are within reach,
the vector travels a random distance (from the exponential
distribution) in a random direction.
(b) Move T. sanguisuga
T. sanguisuga disperse like T. gerstaeckeri except that they
are assumed to prefer raccoons over woodrats, so when dispersing, if there is an occupied den within a days travel the
vector enters the nearest one; if not, then the vector checks
for occupied nests as before.
3. Update hosts
(a) Update woodrats
First the number of woodrats to be born is drawn from a
binomial distribution where n is the number n of woodrats
on the grid and p is the baseline birth probability scaled by
the logistic factor (1 N/K) (where K is the carrying capacity). The given number of newborn woodrats is created, each
randomly placed in a nest with room for it, and its counters
initialized to 0.
Next the number of woodrats to survive is likewise drawn
from a binomial distribution with a given probability. The
number to die is then calculated by subtraction, and the given
number of woodrats are chosen randomly to die.
(b) Update raccoons
Raccoon births and deaths are determined similarly to
woodrats, but with their own probabilities.
4. Update vectors
(a) Update T. gerstaeckeri (TG)
First, the number of natural deaths of adult TG vectors on
the grid is determined (by subtracting from the total number
a random draw from a binomial distribution with given survival probability). For the given number of deaths, the vectors
are selected randomly, their feeding histories are ended and
appended to the global lists of feeding histories (host IDs and
times), and they die.
Next, the surviving vectors and woodrats woodratTG
bite counters are reset to 0, and each vector is examined
concurrently to determine if (a) its hunger counter meets
or exceeds the feeding hunger threshold and (b) there is a
woodrat there. If so, its bite counter, and the corresponding one of a (randomly chosen if not unique) woodrat there
are incremented by 1. Once all adult TG have been polled,
any woodrats with positive TG bite counters are polled

K.E. Yong et al. / Acta Tropica 151 (2015) 2131

25

Table 1
User input parameters (App. refers to Appendix).
Parameter

Denition

Value

Reference

grid size
cell diameter
percent wooded
percent prickly
raccoon dns
woodrat dns
dens/raccoon
nests/woodrat
den infest prop
nest infest prop
vec density
juvprop
nestGprop

Length & width of modelled area


Length & width of a grid cell
Proportion of cells which are raccoon habitat
Proportion of cells which are woodrat habitat
Initial density of raccoons
Initial density of woodrats
Ratio of dens per raccoon
Ratio of nests per woodrat
% of dens initially infested with vectors
% of nests initially infested with vectors
Vector population density
[Initial] proportion of vector pop. which are juveniles
[Initial] proportion of vectors in a woodrat nest which
are T. gerstaeckeri

75 cells
10.4 m
0 or 100
100 or 0
rac
0.08 acre
wr
9.3 acre
9.25
2.1
70.5%
70.5%
vec
128 acre
0.4785
0.4

N/A
Raun (1966), Rogers et al. (2010), Thies et al. (1996)
N/A
N/A
Kribs-Zaleta (2010a)
Kribs-Zaleta (2010a), Raun (1966)
Shirer and Fitch (1970), App.
Merkelz and Kerr (2002)
App.
Eads et al. (1963), Pippin (1970)
Burkholder et al. (1980)
Eads et al. (1963), Pippin (1970), App.
Eads et al. (1963), Pippin (1970), App.

tb

Maximum number of bites a host can tolerate before


leaving due to irritation
# of juvenile vector bites equiv. to an adult bite
Daily probability a raccoon changes den
Daily probability a woodrat changes nest
Maximum distance a raccoon travels in one day
maximum distance a woodrat travels in one day

bites
210 night
(varies)
2.06
0.685
0.25
990 m
27.14 m

Castanera
et al. (2003)

7 days
14 days

Borges et al. (2005), App.


Galvo et al. (2001), Pippin (1970)

36 days

Pippin (1970), App.

mean vec travel

Vector bloodmeal digestion time (min. time btw. bites)


Time since last feeding before a vector disperses to
seek a new host
Maximum adult vector survival time without feeding
(starvation threshold)
Average daily dispersal distance for adult vectors

1000 m

App.

RaccoonCC
WoodratCC
prb
pw
b
prs
pw
s

Density carrying capacity for raccoons


Density carrying capacity for woodrats
Daily prob. that a raccoon births one child
Daily prob. that a woodrat births one child
Daily raccoon survival probability
Daily woodrat survival probability

rac
0.144 acre
wr
21 acre
0.00246
0.00492
0.99891
0.99727

Kribs-Zaleta (2010a)
Kribs-Zaleta (2010a)
Kribs-Zaleta (2010a), App.
Kribs-Zaleta (2010a), App.
Kribs-Zaleta (2010a), App.
Kribs-Zaleta (2010a), App.

psb

0.137

Kribs-Zaleta (2010a), App.

0.355

Kribs-Zaleta (2010a), App.

pssj

Daily probability that an adult T. sanguisuga hatches a


juvenile
Daily probability that an adult T. gerstaeckeri hatches a
juvenile
Daily survival probability for T. sanguisuga juveniles

0.99818

Kribs-Zaleta (2010a), App.

psj

Daily survival probability for T. gerstaeckeri juveniles

0.99586

Kribs-Zaleta (2010a), Martnez-Ibarra et al. (2007), App.

juv starv
psd
g
pd
pssa
g
psa

Maximum juvenile vector survival time w/o feeding


Daily maturation prob. for T. sanguisuga juveniles
Daily maturation prob. for T. gerstaeckeri juveniles
Daily survival probability for T. sanguisuga adults
Daily survival probability for T. gerstaeckeri adults

31.5 days
0.00122
0.00276
0.99999
0.99650

Pippin (1970), App.


Kribs-Zaleta (2010a), Pippin (1970), App.
Pippin (1970), App.
Kribs-Zaleta (2010a), App.
Pippin (1970), App.

juv intense
mr
mw
dr
dw
tw
th
ts

g
pb

concurrently to determine host irritability. For those


woodrats, the number of adult T. sanguisuga in that nest with
hunger counter meeting or exceeding the feeding hunger
threshold is added to the total number of juvenile vectors
in the nest rescaled for relative intensity and frequency of
bite; if the sum exceeds the host irritability threshold, then
the host is considered irritated, and the number of successful
adult TG feeding is set at the host irritability threshold multiplied by the proportion of desired bites which were adult TG
(the product is rounded down to a whole number). The number of unsuccessful TG biters (calculated by subtraction) is
then determined, and that many attempted biters then have
their bite counters reset back to zero.
Finally, the adult vectors are again polled concurrently: the
successful biters have their hunger counters reset and their
feeding histories updated, and those who did not feed have
their hunger counters incremented, and those whose hunger
counters meet the starvation threshold have their feeding
histories retired and appended to the global lists, and then
they die.

Castanera
et al. (2003), Pippin (1970), App.
Rogers et al. (2010), Shirer and Fitch (1970)
App.
Rogers et al. (2010), Shirer and Fitch (1970)
Raun (1966), Rogers et al. (2010)

(b) Update T. sanguisuga (TS)


First, natural death of adult vectors is determined, as for
TG but with a different survival probability. Next, host and
vector TS bite counters are reset. Then the vectors are polled
concurrently to see if their hunger counters meet or exceed
the feeding threshold. If a woodrat is present there, the
vector tries to bite it (host and vector bite counters are incremented); if a raccoon is present there, the vector tries to bite
it (order is irrelevant since woodrats and raccoons do not
share habitat). Following this, woodrats are polled concurrently to see which have positive TS bite counters. For those
that do, their host irritability is determined, by comparing
to the irritability threshold the number of attempted bites
(calculated as the sum of the woodrats two species-specic
vector bite counters plus the total number of juvenile vectors
there, rescaled as before). For irritated hosts, the number of
successful TS biters is calculated as the irritability threshold multiplied by the proportion of attempted bites made
by adult TS. The appropriate number of unsuccessful biters
are chosen at random and their bite counters reset. Then

26

K.E. Yong et al. / Acta Tropica 151 (2015) 2131

raccoons are similarly checked for irritability (though they


have no attempted TG bites to tally). Finally, each adult TS
is polled concurrently either to reset or to increment their
hunger counter, and those who did not feed and now have
a hunger counter at the starvation threshold die as detailed
for TG.
5. Update homes
(a) Update nests (juvenile vectors)
Woodrat nests are polled concurrently: if unoccupied,
their host absence counters are incremented, and if the host
absence counter meets the juvenile starvation threshold, the
juveniles in the nest die. If occupied, the host absence counter
is reset to zero. Next, the number of juvenile vectors of each
species in that nest is incremented by a draw from a binomial
distribution with n the number of adult vectors of the given
species in the nest, and p the species birth rate. Next, the
number of juvenile vectors of each species in the nest surviving to the next time step is determined by a draw from a
binomial distribution with given probability. Then the number of juvenile vectors of each species there maturing into
adults is also determined by a draw from a binomial distribution with given probability; the corresponding adult vectors
are created, with properties appropriately initialized, and
the number of juveniles there of that type is decremented
correspondingly.
(b) Update dens (juvenile vectors)
The host absence counter and juvenile vector population
of each raccoon den are updated concurrently in a similar
way as for nests, but with different probabilities and no T.
gerstaeckeri.
Parameter estimates given in Table 1 are discussed in more
detail in Appendix, and in a separate literature review (Rogers et al.,
2010).
3. Results
Initial analysis was made from 20 simulations for each of the
four scenarios (raccoon or woodrat cycles; host irritability threshold tb = 2 or 10 bites per host per night), and then extended to 100
simulations. In all simulations the host populations held steady
near their initial [equilibrium] values (and will therefore not be further discussed), but the vector densities varied substantially over
time. Juvenile biting rates were computed (and are here reported)
in units of adult-equivalent bites, and should be multiplied by the
equivalence factor of 2.06 given in Table 1 to get the number of
juveniles per host or vector per night that fed.
Raccoons: The relatively low host density (compared to
woodrats) coupled with a low host irritability threshold severely
curtails vectors ability to feed, with mass starvation dropping adult
vector density from 10,000 over the entire grid to a steady mean
(across simulations) of just 662.5 (tb = 2) or 1630 (tb = 10). The juvenile TS population, which starts out in growth mode, undergoes
a similar reset, dropping an order of magnitude from a high of
nearly 20,000 to a nal mean of 2023 (tb = 2) and 3900 (tb = 10).
The resulting per capita biting rates for juvenile TS fall during the
initial growth period but recover when the adult population undergoes starvation, thereafter leveling out at a fairly steady 0.000978
(tb = 2) and 0.003654 (tb = 10) adult bites per juvenile per day. The
per capita adult TS biting rates, on the other hand, begin low and
then rise when the juvenile population begins to starve as well,
settling at a markedly higher 0.03735 and 0.04600 adult bites per
adult per day. From the hosts perspective, meanwhile, we see three
phases: if tb = 2, the average host receives 1.51.75 bites per night
(less than the maximum of 2 since raccoons spend some nights

in uninfested dens), rst split 40/60 juveniles/adults, then mostly


juveniles, and nally nearly all adults. Graphs for all of these trends
are given as an example in Fig. 1 (for tb = 2; the overall shape is similar for tb = 10). The fact that the total number of bites an average host
received per night for tb = 10 remained well above 9 throughout
the reset period underlines the key role played by host irritability
in controlling vector density, and the average interval between
bites for a given vector, which may be calculated as the reciprocal of the corresponding per capita rate, offers a key measure of
survivability when compared with the starvation threshold (from
Table 1, 36 days for adults and 31.5 days for juveniles): The interval
remains well above 100 days for juveniles over time for both tb = 2
and tb = 10, indicating that the relative scarcity of raccoons and the
consequent infrequency of their visits to each den are not enough
to sustain substantial juvenile populations (caveat: this model does
not take into account the visits to these same dens made by other
hosts such as opossums and skunks), while for adult vectors the
mean feeding interval drops from over 120 days to under 27 days for
both tb values, indicating arrival at a sustainable level. In summary,
under such difcult feeding conditions vector densities inevitably
drop, with adult bites dominating juvenile feeding at the nal
levels.
Woodrats: The relative abundance of woodrats meant in simulations that regardless of irritability threshold they visited infested
nests often enough to prevent the mass juvenile starvation that
occurred in raccoon dens, with the result that the juvenile populations of both TS and TG showed steady growth (although it was
concave down for tb = 2 and concave up for tb = 10). It should be
noted that this is in part a limitation of the model, which for
computational efciency calculated [successful] juvenile bites but
determined juvenile starvation on a per-nest, not per-juvenile,
basis. Under low irritability threshold (tb = 2), the adult vector densities for both species still underwent a starvation reset between
23 and 37 days; the TS mean dropped from 6053 to a more or less
level 1540 vectors across the grid, while the TG density quickly
returned to growth mode which continued past 100 days. With
hosts both plentiful and more bite-tolerant (tb = 10), the simulations showed consistent and accelerating growth for juvenile and
adult vector densities of both species, suggesting that the true value
of tb is lower. Per capita biting rates for tb = 2 for both juvenile
and adult TS declined steadily over time, with respective means
of 0.02680 and 0.009870 bites per day per vector over the 100
days, both too low to beat starvation, suggesting a struggle at
least in the short term for TS under these conditions. The TG
rates leveled off for both juveniles and adults at tb = 2, but at even
lower levels, 0.006354 and 0.00207 bites per vector per day, suggesting that most vectors do starve, compensating for the high
reproduction rate. Mean per-host juvenile biting rates were 0.5333
(TS) and 0.3003 (TG) per night over 100 days, and around 0.01
for adults. The per capita biting rates for a high biting threshold
(tb = 10), meanwhile, declined over time due to competition with
growing numbers, but at rates of 0.06445 and 0.0678 bites per
vector per night for TS (resp. juveniles and adults) and 0.04261
and 0.04873 for TG, all well above the critical rate to avoid starvation. Although the juvenile bites dominate the adult bites in
absolute number due to their higher densities, the totals per host
per night still fall well short of the bite ceiling, suggesting that if
woodrats were able to tolerate so many bites per night their associated vector densities would be markedly higher. For comparison
purposes, the mean adult vector densities for both species and both
irritability thresholds are graphed in Fig. 2 (note vertical scales
differ).
Lastly, we consider the outcomes of the adult vector dispersals
motivated by hunger. Fig. 3 presents the six histograms showing the
frequency of duration of dispersals from 1 to 23 days (with 23 days,
the last bar, representing starvation); underneath each histogram

K.E. Yong et al. / Acta Tropica 151 (2015) 2131

Mean juvenile TS population

Mean juvenile TS bites juv day

20 000

0.0008

15 000

0.0006

10 000

0.0004

5000

0.0002
20

40

60

80

Mean juvenile TS bites host day


1.5
1.0
0.5

20

100

27

40

60

80

100

20

Mean adult TS bites adult day

Mean adult TS population


10 000

0.04

8000

0.03

40

60

80

100

Mean adult TS bites host day


1.5

6000

1.0

0.02

4000
0.5

0.01

2000
20

40

60

80

100

20

40

60

80

100

20

40

60

80

100

Fig. 1. Graphs of mean (top) juvenile and (bottom) adult TS densities and biting rates over time, for 100 days, on raccoons with tb = 2 bites/raccoon/day: left, vector densities;
center, per-vector biting rates; right, per-host biting rates (rates in adult bites per day).

Fig. 2. Graphs of mean adult vector densities over time for 100 days associated with woodrats with (top) tb = 2, (bottom) tb = 10: left, TS; right, TG.

is given the proportion of failed dispersals and mean duration of


successful dispersals. As can be seen in the gure, although the
increase in host tolerance reduces both the proportion of failure
and the mean length of successful dispersal, it has a much more
dramatic effect on vectors associated with woodrats than on those
associated with raccoons. In both vector species, starvation prior to
nding a new host becomes extremely rare, and it takes on average
much less time to nd one. This is consistent with the signicant
difference seen in vector densities and biting rates as woodrat tolerance increases, relative to the impact of tb on raccoon-associated
vectors.
A preliminary look at the results of the simulations performed
for this study suggests that in practice dispersal in search of hosts
is not only common (consistent with observations in the literature, e.g., Pippin, 1970) but usually does result in switching to a
new host (rather than nding the previous one in a new location).
Although this model ignores territoriality in hosts, which may limit
vectors ability to disperse far enough to switch, the fact that the
overwhelming majority of bloodmeals involve a different host than
the previous bloodmeal helps explain the spread of parasites such
as T. cruzi within a host population.

4. Discussion
The preliminary results reported in this manuscript offer both
quantitative and qualitative descriptions of vectorhost interactions that are impractical to obtain via eld studies yet critical
to understanding the sylvatic transmission of parasites like T.
cruzi within and between reservoir cycles. To our knowledge, no
other studies of any kind have tried to estimate actual sylvatic
hostvector contact rates in the eld. In addition to offering baseline estimates of contact rates, the simulations made using the
agent-based model described herein conrm host irritability as a
main limiting factor (in conjunction with host density) for vector
population density, which in turn affects the contact rates per vector or per host. The relative scarcity of hosts like raccoons constrains
vector growth even for high bite tolerance, while the relative abundance of hosts like woodrats allows host irritability to play a crucial
role in the growth or decline of vector densities, as well as more
specically the success or failure (and duration) of vector dispersal,
which is the root of the host-switching that spreads the parasite.
Simulations of the transmission cycle involving raccoons and
T. sanguisuga showed that starvation/host unavailability would

28

K.E. Yong et al. / Acta Tropica 151 (2015) 2131

TS/raccoon

TS/woodrat

TG/woodrat

0.07

0.05

0.06

0.15

0.04

Irritability
threshold 2

0.05
0.10

0.03

0.03

0.02

0.02

0.05

0.01

0.01
5

10

15

20

10

15

20

10

15

20

%starve: 16.36%

%starve: 3.98%

%starve: 4.52%

MDSD: 11.08 days

MDSD: 10.32 days

MDSD: 11.13 days


0.12

0.15

Irritability
threshold 10

0.04

0.25

0.10
0.05

10

15

20

0.10

0.20

0.08

0.15

0.06

0.10

0.04

0.05

0.02
5

10

15

20

10

15

20

%starve: 14.45%

%starve: 0.13%

%starve: 1.31%

MDSD: 10.41 days

MDSD: 4.374 days

MDSD: 7.603 days

Fig. 3. Frequency of duration of adult vector dispersals from 1 to 23 days for each vector/host combination (the last bar represents starvation/failure) and tb = 2 vs. 10, along
with failure proportion (%starve) and mean duration of successful dispersals (MDSD).

signicantly reduce juvenile and adult vector densities well below


the densities reported in association with woodrats, regardless of
host irritability threshold; this suggests that the actual vector densities in raccoon dens are much lower, although further work is
needed to estimate true vector densities when other hosts such
as opossums and skunks which sometimes use the same dens as
raccoons (not necessarily at the same time) are included.
The decline in vector densities when woodrats bite tolerance
is limited to 2 per night, combined with their growth when that
tolerance is raised to 10, suggests that woodrats actual tolerance
is somewhere in between the two values, and further ne tuning
can produce an estimate of the irritability threshold consistent with
observed densities. The difference in densities and biting rates for T.
sanguisuga and T. gerstaeckeri in woodrat nests is also worth noting,
with values for the latter species (TG) consistently rising over time
in situations where those for the former species (TS) continue a
decline despite being more common in nests.
Quantitatively, then, the per-host contact rates approach the
host irritability threshold, except when both that threshold and
host density are high; likewise, when vector densities reach sustainable levels, the per-vector contact rates are on average just
high enough to avoid starvation, but not high enough to avoid
the need for dispersal. The exception was the scenario with
high host density (woodrats) and high bite tolerance (tb = 10), for
which the TS mean biting rates hovered around the frequency
(14 days)1 matching the dispersal threshold while TG mean
biting rates were approximately (22 days)1 over the 100-day
simulations.
Further work using this model will examine in detail the rates
at which individual vectors change hosts, as well as the rates of
host type switching, the mechanism for spreading parasites across
cycles.
Acknowledgments
KY was supported by NSF grant DMS-0946431; AM and CK
acknowledge the support of NSF grant DMS-1020880. The authors

also acknowledge the work of their student coauthors in developing


(Rogers et al., 2010) a preliminary version of the present study.
Appendix. Parameter estimation
A.1. Host densities and homes
As noted in the main text, cell diameters were set at 10.4 m in
the model since this reects (Rogers et al., 2010) the approximate
minimum distance between woodrat nests in one study (Merkelz
and Kerr, 2002), the mean distance between nests in another study
(Thies et al., 1996), and the mean distance traveled daily by a
woodrat in a third (Raun, 1966). With a 75 75 grid, the resulting area is about 60 ha or 150 acres. Population densities can be
multiplied by the proportion of this area which contains suitable
habitat, in order to gure absolute population sizes.
Woodrat population density estimates vary by more than an
order of magnitude, cf. reviews in (Kribs-Zaleta, 2010a; Rogers
et al., 2010). (Some of the variation in published estimates may be
explained by seasonal uctuations.) Here we use 9.3 woodrats/acre
(23 woodrats/ha), which is both the mean found across all studies considered in (Kribs-Zaleta, 2010a) and the density found in
(Raun, 1966), although the cross-study mean found in (Rogers et al.,
2010) is lower, 5.79 woodrats/acre. Likewise the studies reviewed
in (Kribs-Zaleta, 2010a) found raccoon densities varying from 0.002
to 0.083 raccoons/acre in sylvatic sites and up to 0.5 raccoons/acre
in peridomestic settings and national parks in the U.S.; we here use
the average of 0.08 raccoons/acre (0.2 raccoons/ha) in sylvatic sites
given in (Kribs-Zaleta, 2010a).
The southern plains woodrat, Neotoma micropus, lives primarily
in shortgrass landscapes dominated by prickly pear cactus, where
they make their nests (and which serves as primary food source)
(Box, 1959; Braun and Mares, 1989; Conditt and Ribble, 1997; Raun,
1966). These woodrats live generally alone in single nests, although
males may range over multiple nests, some of them shared with
females (Braun and Mares, 1989; Merkelz and Kerr, 2002; Raun,
1966). An extensive 20-month study by Merkelz and Kerr found

K.E. Yong et al. / Acta Tropica 151 (2015) 2131

some nests used by a succession of 5 or more different rats during


the study (Merkelz and Kerr, 2002). They also found a roughly 1:1
sex ratio, with 24% of recaptured females and 69% of recaptured
males found at multiple nests. Overall they tagged 180 rats at 86
nests, for a nests/woodrat ratio of 2.1.
The probability of a woodrat changing nests from one day to
the next (for reasons other than host irritability) was estimated
at 1/4 by supposing that half of woodrats are female and do not
normally change nests, while the other half are male and change
nests on average every 2 days. We found no empirical estimates in
the literature for this quantity. The only major study of woodrat nest
use (Merkelz and Kerr, 2002) made 239 captures at 86 nests over
2292 trap-nights during 20 months, so each nest was checked an
average of 2292/86=26.65 times, for an average frequency of 1.33
times per month, not often enough to derive estimates for daily
change rates.
Shirer and Fitch (1970) studied the movements of four adult
raccoons, observed respectively in 4, 7, 12, and 14 different locations, for an average of 9.25 dens per raccoon. They also studied
four juvenile raccoons but did not disentangle the number of locations recorded for each. A literature review did not nd any data
on the proportion of raccoon dens (or opossum or skunk, which
often share dens, not necessarily simultaneously) infested with triatomines, so we use the proportion for woodrat nests as a proxy.
A.2. Vector density
Eads et al. (1963) dismantled 80 woodrat nests and found 390
Triatoma in 58 of 80 nests (72.5%): by life stage, 372 juveniles and
18 adults total, which gives a paltry 6.4 (mean) juveniles and 0.31
adults per infested nest, and 4.875 total vectors per overall nest; by
species, 226 T. sanguisuga, 133 T. gerstaeckeri, 31 Triatoma neotomae
by species (T. neotomae specialize in woodrats but are rare in Texas,
more common in Mexico), which gives T. sanguisuga a 1.7:1 advantage over T. gerstaeckeri (63.0%/37.0% excluding T. neotomae) in
nests. On rst glance the small numbers might seem to contradict
Burkholder et al.s estimated density (it would take 26.25 nests per
acre to replicate the observed 128 bugs/acre), but in fact Eads et al.
observe that what it really means is that adults are frequently out
foraging and not in the nest all the time. So we should expect to see
lots of dispersal in our simulations.
Pippin (1970) searched 142 nests and found 229 bugs in 85
nests (68.5%): by life stage, 213 nymphs and 16 adults total, or
2.5 juveniles and 0.19 adults per infested nest (range 122 total),
1.612 total vectors per overall nest; by species, 123 + 9 =132 T. sanguisuga (nymphs + adults), 90 + 7 =97 T. gerstaeckeri, for a 1.36:1
ratio T. sanguisuga:T. gerstaeckeri (57.6%/42.4%). The study includes
the following passage supporting the notion of widespread adult
dispersal:
The bugs were rarely found in the rats nests. The preferred
habitat appeared to be in cracks and crevices in tunnels and runways leading to the nest or in debris covering the nest. Nymphs
were occasionally found under logs and debris as far as 90 m
from the nest of any known host. The condition of most of the
bugs indicated they had not fed for some time. This was true
even when the bugs were within 60 cm of an occupied nest.
We take the averages from these two papers of 70.5% for nest
infestation proportion and 60%/40% T. sanguisugaT. gerstaeckeri ratio. Our estimates of 128 vectors/acre, 9.3 woodrats/acre,
2.1 nests/woodrat and 70.5% nest infestation, however, produce
an expected average 9.3 vectors per infested nest, compared to
Eads et al.s 6.71 and Pippins 2.69. Since the latter two studies
concentrated around nests, we interpret the difference to indicate
adults dispersed away from nests, and use the higher gure to seed
the model (all vectors begin in nests but adults may eventually

29

disperse to match observed data). Thus of 9.3 vectors per average


infested nest, we suppose a mean (of 6.4 and 2.5) 4.45 juveniles
and 9.3 4.45 = 4.85 adults, making juveniles represent 47.85% of
the vector population.
A.3. Host irritability, vector feeding and dispersal

In studying host irritability, (Castanera


et al., 2003) estimated
bites of the rst through third instars to contribute only 1/10 as
much to host irritability as an adult vectors bite, and bites of fourth
and fth instars to contribute 0.8 times as much. (In comparison,
Pippin found Pippin, 1970, Table 8 that the volume of blood taken
per bloodmeal increased by about 1/2 an order of magnitude per
instar stage for T. sanguisuga, and half as much for T. gerstaeckeri.)
They also estimated host irritability at between 2 and 10 bites per
day per host (described as vector density but given in units of per
day). Their results indicated that differentiating between nymphs
and adults contributions to host irritability was crucial to reproducing observed patterns in experimental data (although precise
differences between nymph stages were not), and that the host
irritability threshold does play an important role in vector density.
Here, we use Pippins (1970, Table 5) data on mean duration of
each instar stage for T. sanguisuga under varying (1830 C) and
constant (27 C) temperatures, and T. gerstaeckeri under varying
temperatures, to estimate 45% of juveniles in the rst three instars
and 55% in the last two, together with relative irritation factors

of 0.1 and 0.8 from Castanera


et al. (2003), to estimate an overall irritation factor of 0.485 for juveniles relative to adult bites, or
2.06 juvenile bites irritation-equivalent to 1 adult bite. (Data for
T. gerstaeckeri under constant temperature exhibited a markedly
different pattern, with only 24.3% of a juveniles life spent in the
rst three instars.) In comparison, a similar stagewise average of
bloodmeal size weighted by mean stage duration using Pippins
data produces an equivalence factor of 1.24 or 1.56 for T. sanguisuga
and 1.77 or 2.48 for T. gerstaeckeri juvenile bloodmeals to 1 adult
bloodmeal.
Although we found no direct studies of preferred feeding frequency for T. sanguisuga or T. gerstaeckeri, a 7-day interval is
consistent with results from the literature for other triatomine
species. Borges et al. reported a feeding frequency of once per week
for Triatoma brasiliensis (Borges et al., 2005). Other studies reported
mean times between bloodmeals ranging from 6.24 to 10.74 days
for some South American triatomines (Canals et al., 1999), 49 days
for some domestic South American triatomines (Catal, 1991), or
1.73.4 days for peridomestic populations of Triatoma infestans in
Argentina during spring and summer (Lpez et al., 1999). Ceballos
et al. (2005) wrote of Triatoma infestans, Mean daily feeding rates
decreased signicantly from 34% in spring to 23% in summer to 14%
and 18% in fall and winter, respectively; mean feeding intervals
were 2.9, 4.3, 7.0 and 5.6 days, respectively.
Pippin (1970) captured 698 T. gerstaeckeri using outdoor light
traps at night and found that only 2 had recently fed. He wrote that
their general appearance was similar to bugs which had not fed for
1421 days. This is consistent with Galvo et al.s study of Triatoma
infestans and the closely related Triatoma melanosoma (Galvo et al.,
2001), which found that the peak vector dispersal time was 14
days after feeding. The ten unfed females which Pippin then tested
for longevity survived from 4 to 22 days with a mean of 13. If we
suppose that bugs within less than 4 days of starvation were not
attracted to the traps, this allows up to at least 22 days survival postdispersal before starvation. (Note that the data in Pippins Table
12 for survival on one feeding, with means well over 100 days for
adults of both species, were collected under eld conditions during
the winter, beginning on November 1, 1966; the discrepancy with
the lower values here is consistent with other results from Pippin
which suggest that the bugs go dormant in the winter. Whether

30

K.E. Yong et al. / Acta Tropica 151 (2015) 2131

or not it is true diapause remains to be studied.) From all this we


assume that dispersal initiates 14 days after feeding, and starvation
occurs 36 days after feeding.
Pippin studied starvation longevity of adults in the lab, and of
adults and instars of all stages in the eld over the winter (beginning November 1); for T. sanguisuga there was a consistent factor
of 4.5 between lab longevity and outdoor winter longevity for both
sexes of adult, so applying the same factor to the winter survival of
nymphs, the average extrapolated starvation longevity for juveniles
is 31.5 days, consistent with the estimate above for adults.
A literature review found no studies with information on dispersal distances for T. sanguisuga or T. gerstaeckeri. Schoeld,
Lehane and colleagues conducted several studies on dispersal of
South American species including T. infestans and T. sordida and
found, in general, that of the bugs they released to disperse, about
half went nowhere (remaining 5 m or less from the release spot)
and the other half ew 50 m or more (most of them ew 100 m
or more and were lost to the researchers); in one case a vector
was found to have traveled 1350 m (Lehane and Schoeld, 1978;
Schoeld et al., 1991, 1992). Another set of researchers studying Triatoma dimidiata in Mexico (Ramirez-Sierra et al., 2010) bounded its
single-ight distance at between 95 m and 120 m, although the context of dispersal to houses in a small village from its periphery may
not represent the vectors maximum capacity. Aware of the possible effect of dispersal distance, we take a mean of 1000 m, but note
that preliminary simulations using much lower values resulted in
near-universal failure to nd new hosts.
A.4. Rates converted to probabilities
The vector reproduction rates (per capita per year) were calculated by taking the expressions given in Kribs-Zaleta (2010a, Supp.)
and removing the 60% survival-to-maturity probability. Likewise
vector longevities in the same source were broken down into timeto-maturation and adult lifespan. For each vector species, the 60%
survival-to-maturity proportion was then converted into a juvenile
j
mortality rate j by setting  +m
= 0.6 (where m is the maturaj

tion rate), whence j = 1.5m. Host reproduction and mortality rates


were taken directly from Kribs-Zaleta (2010a). All base rates are
given in Table A1.
To convert these rates into daily probabilities, rst the units
were converted to days1 (by dividing per-year rates by 365.24).
Then for each rate, say , the probability of a given transition/event occurring or not occurring was determined using a
simple differential equation dx
= x, whence the proportion of
dt
the population which has not made the transition after T units of
time is x(T)/x(0) = eT , and the proportion that has made the transition (e.g., reproducing, maturing, dying) is then 1 eT . (Survival
is thus dened in terms of not dying.) These proportions were used

Table A1
Conversion of baseline event rates from literature into probabilities of the events
occurring in a single day (T.s. = T. Sanguisuga, T.g. = T. gerstaeckeri).
Parameter

Base rate (from lit.)

Probability (1 day)

Raccoon birth
Woodrat birth
Raccoon death
Woodrat death
T.s. birth
T.g. birth
T.s. juv. death
T.g. juv. death
T.s. maturation
T.g. maturation
T.s. adult death
T.g. adult death

0.9/year
1.8/year
0.4/year
1/year
54/year
160/year
1.5/(2.25 year)
1.5/(361.9 day)
1/(2.25 year)
1/(361.9 day)
1/(525 day)
1/(285.2 day)

0.00246
0.00492
0.00109
0.00273
0.137
0.355
0.00182
0.00414
0.00122
0.00276
0.00190
0.00350

for probabilities, with  in days1 and T = 1 day. This derivation


corresponds to the probabilities in a Poisson waiting process with
parameter .
Finally, since starvation and natural mortality [due to causes
other than starvation] are modeled separately in the ABM, we
adjusted the vector survival probabilities by increasing them to
offset the observed starvation rates, which were remarkably consistent across simulations. T. sanguisugas daily per capita starvation
rate in the model was observed at 0.00189, accounting for nearly
all adult deaths (total mortality estimated at 0.00190, cf. Table A1),
so the adult daily survival probability [due to other causes] rose to
0.99999.
References
Barbu, C., Dumonteil, E., Gourbire, S., 2010. Characterization of the dispersal of
non-domiciliated Triatoma dimidiata through the selection of spatially explicit
models. PLoS Negl. Trop. Dis. 4 (8), e777, http://dx.doi.org/10.1371/journal.pntd.
0000777.
Borges, E.C., Dujardin, J.P., Schoeld, C.J., Romanha, A.J., Diotaiuti, L., 2005. Dynamics
between sylvatic, peridomestic and domestic populations of Triatoma brasiliensis (Hemiptera: Reduviidae) in Ceara state, northeastern Brazil. Acta Trop. 93,
119126.
Box, T.W., 1959. Density of plains wood rat dens on four plant communities in South
Texas. Ecology 40 (4), 715716.
Braun, J.K., Mares, M.A., 1989. Neotoma micropus. Mamm. Species 330, 19.
Burkholder, J.E., Allison, T.C., Kelly, V.P., 1980. Trypanosoma cruzi (Chagas) in invertebrate, reservoir, and human hosts of the Lower Rio Grande Valley of Texas. J.
Parasitol. 66 (2), 305311.
Canals, M., Sols, R., Tapia, C., Ehrenfeld, M., Cattan, P.E., 1999. Comparison of some
behavioral and physiological feeding parameters of Triatoma infestans Klug, 1834
and Mepraia spinolai Porter, 1934, vectors of Chagas disease in Chile. Mem. Inst.
Oswaldo Cruz 94 (5), 687692.

Castanera,
M.B., Aparicio, J.P., Grtler, R.E., 2003. A stage-structured stochastic model
of the population dynamics of Triatoma infestans, the main vector of Chagas
disease. Ecol. Model. 162, 3353.
Catal, S.S., 1991. The biting rate of Triatoma infestans in Argentina. Med. Vet. Entomol. 5, 325333.
Ceballos, L.A., Vazquez-Prokopec, G.M., Cecere, M.C., Marcet, P.L., Grtler, R.E., 2005.
Feeding rates, nutritional status and night dispersal potential of peridomestic
populations of Triatoma infestans in rural northwestern Argentina. Acta Trop. 95
(2), 149159.
Conditt, S.A., Ribble, D.O., 1997. Social organization of Neotoma micropus, the southern plains woodrat. Am. Midl. Nat. 137 (2), 290297.
DeAngelis, D.L., Gross, L.J., 1992. Individual-based Models and Approaches in Ecology: Populations, Communities and Ecosystems. Taylor and Francis.
Devillers, H., Lobry, J.R., Menu, F., 2008. An agent-based model for predicting the
prevalence of Trypanosoma cruzi I and II in their host and vector populations. J.
Theor. Biol. 255, 307315.
Eads, R.B., Trevino, H.A., Campos, E.G., 1963. Triatoma (Hemiptera: Reduviidae)
infected with Trypanosoma cruzi in south Texas wood rat dens. Southwest. Nat.
8 (1), 3842.
Gaff, H., 2011. Preliminary analysis of an agent-based model for a tick-borne disease.
Math. Biosci. Eng. 8 (April (2)), 463473, http://dx.doi.org/10.3934/mbe.2011.8.
463.
Galvo, C., da Silva Rocha, D., Jurberg, J., Carcavallo, R., 2001. Incio da Atividade de
Vo em Triatoma infestans (Klug, 1834) e T. melanosoma Martnez, Olmedo &
Carcavallo, 1987 (Hemiptera, Reduviidae). Mem. Inst. Oswaldo Cruz 96 (January
(1)), 137140, Online at: http://memorias-old.ioc.ocruz.br/961/4050.pdf
Galvo, V., Garcia Vivas Miranda, J., Ribeiro-dos-Santos, R., 2008. Development of a
two-dimensional agent-based model for chronic chagasic cardiomyopathy after
stem cell transplantation. Bioinformatics 24 (18), 20512056, http://dx.doi.org/
10.1093/bioinformatics/btn362.
Grimm, V., Railsbeck, S.F., 2005. Individual-based Modeling and Ecology. Princeton
University Press.
Grimm, V., Berger, U., Bastiansen, F., Eliassen, S., Ginot, V., Giske, J., Goss-Custard, J.,
Grand, T., Heinz, S.K., Huse, G., Huth, A., Jepsen, J.U., Jrgensen, C., Mooij, W.M.,
Mller, B., Peer, G., Piou, C., Railsback, S.F., Robbins, A.M., Robbins, M.M., Rossmanith, E., Rger, N., Strand, E., Souissi, S., Stillman, R.A., Vab, R., Visser, U.,
DeAngelis, D.L., 2006. A standard protocol for describing individual-based and
agent-based models. Ecol. Model. 198, 115126.
Grimm, V., Berger, U., DeAngelis, D.L., Polhill, J.G., Giske, J., Railsback, S.F., 2010. The
ODD protocol: a review and rst update. Ecol. Model. 221, 27602768.
Guerenstein, P.G., Lazzari, C.R., 2009. Host seeking: how triatomines acquire and
make use of information to nd blood. Acta Trop. 110, 148158.
Kjos, S.A., Snowden, K.F., Olson, J.K., 2009. Biogeography and Trypanosoma cruzi
infection prevalence of Chagas disease vectors in Texas, USA. Vector-Borne
Zoonotic Dis. 9 (1), 4150.
Kribs-Zaleta, C.M., 2004. To switch or taper off: the dynamics of saturation. Math.
Biosci. 192 (2), 137152.
Kribs-Zaleta, C.M., 2006. Vector consumption and contact process saturation in sylvatic transmission of T. cruzi. Math. Popul. Stud. 13 (3), 135152.

K.E. Yong et al. / Acta Tropica 151 (2015) 2131


Kribs-Zaleta, C.M., 2009. Sharpness of saturation in harvesting and predation. Math.
Biosci. Eng. 6 (4), 719742.
Kribs-Zaleta, C.M., 2010a. Estimating contact process saturation in sylvatic transmission of Trypanosoma cruzi in the U.S. PLoS Negl. Trop. Dis. 4 (4), e656.
Kribs-Zaleta, C.M., 2010b. Alternative transmission modes for Trypanosoma cruzi.
Math. Biosci. Eng. 7 (3), 661676.
Lpez, A., Crocco, L., Morales, G., Catal, S., 1999. Feeding frequency and nutritional
status of peridomestic populations of Triatoma infestans from Argentina. Acta
Trop. 73, 275281.
Lazzari, C.R., Pereira, M.H., Lorenzo, M.G., 2013. Behavioural biology of Chagas disease vectors. Mem. Inst. Oswaldo Cruz 108 (Suppl. I), 3447.
Lehane, M.J., Schoeld, C.J., 1978. Measurement of speed and duration of ight by
triatomine bugs. Trans. R. Soc. Trop. Med. Hyg. 72, 438.
Martnez-Ibarra, J.A., Alejandre-Aguilar, R., Paredes-Gonzlez, E., Martnez-Silva,
M.A., Solorio-Cibrin, M., Nogueda-Torres, B., Trujillo-Contreras, F., NoveloLpez, M., 2007. Biology of three species of North American Triatominae
(Hemiptera: Reduviidae: Triatominae) fed on rabbits. Mem. Inst. Oswaldo Cruz
102, 925930.
Merkelz, Robert, Kerr, Sara F., 2002. Demographics, den use, movements, and
absence of Leishmania mexicana in southern plains woodrats (Neotoma micropus). Southwest. Nat. 47 (1), 7077.
Noireau, F., Dujardin, J.P., 2010. Biology of Triatominae. In: Telleria, J., Tibayrenc,
M. (Eds.), American Trypanosomiasis Chagas Disease: One Hundred Years of
Research. Elsevier, Boston, pp. 149168.
Pelosse, P., Kribs-Zaleta, C.M., 2012. The role of the ratio of vector and host densities
in the evolution of transmission modes in vector-borne diseases. The example
of sylvatic Trypanosoma cruzi. J. Theor. Biol. 312, 133142, http://dx.doi.org/10.
1016/j.jtbi.2012.07.028.
Pippin, W.F., 1970. The biology and vector capability of Triatoma sanguisuga texana Usinger and Triatoma gerstaeckeri compared with Rhodnius prolixus. J. Med.
Entomol. 7 (1), 3045.
Rabinowitz, A.R., Pelton, M.R., 1986. Day-bed use by raccoons. J. Mammal. 67 (4),
766769.

31

Ramirez-Sierra, M.J., Herrera-Aguilar, M., Gourbire, S., Dumonteil, E., 2010. Patterns
of house infestation dynamics by non-domiciliated Triatoma dimidiata reveal a
spatial gradient of infestation in rural villages and potential insect manipulation
by Trypanosoma cruzi. Trop. Med. Int. Health 15 (1), 7786.
Raun, G.G., 1966. A population of woodrats (Neotoma micropus) in southern Texas.
Texas Mem. Mus. Bull.
Rogers, C., Ngan Tran Nguy, K., Amajor, O.A., Seaquist, T., Crawford, B., Mubayi, A.,
Kribs-Zaleta, C., Oyediran, O., 2010. Daily Behavior of Trypanosoma cruzi Hosts
and Vectors in Texas: An Agent-based Modeling Approach in NetLogo, Mathematics Department Technical Report 2010-18. University of Texas at Arlington,
Online at http://www.uta.edu/math/preprint/rep2010 18.pdf
Schoeld, C.J., Williams, N.G., de, T.F., Marshall, C., 1986. Density-dependent
perception of triatomine bug bites. Ann. Trop. Med. Parasitol. 80 (3),
351358.
Schoeld, C.J., Lehane, M.J., McEwen, P., Catala, S.S., Gorla, D.E., 1991. Dispersive
ight by Triatoma sordida. Trans. R. Soc. Trop. Med. Hyg. 85, 676678.
Schoeld, C.J., Lehane, M.J., McEwen, P., Catala, S.S., Gorla, D.E., 1992. Dispersive
ight by Triatoma infestans under natural climatic conditions in Argentina. Med.
Vet. Entomol. 6, 5156.
Schoeld, C.J., 1982. The role of blood intake in density regulation of populations
of Triatoma infestans (Klug) (Hemiptera: Reduviidae). Bull. Entomol. Res. 72,
617629.
Shirer, H.W., Fitch, H.S., 1970. Comparison from radio tracking of movements and
denning habits of the raccoon, striped skunk, and opossum in northeastern
Kansas. J. Mammal. 51, 491503.
Slimi, R., El Yacoubi, S., Dumonteil, E., Gourbire, S., 2009. A cellular automata model
for Chagas disease. Appl. Math. Model. 33, 10721085.
Thies, Kathleen M., Thies, Monte L., Caire, W., 1996. House construction by the southern plains woodrat (Neotoma micropus) in southwestern Oklahoma. Southwest.
Nat. 41 (2), 116122.
Wilensky, U., NetLogo, Center for Connected Learning and Computer-Based Modeling, Northwestern University, Evanston, IL, 1999. http://ccl.northwestern.edu/
netlogo